In one year, generative AI adoption in corporate legal departments nearly doubled. In 2025, 44% of these departments were already using some form of generative AI. In 2026, the figure is 87%. Another number did not keep up with that curve. The share of departments with a formal technology roadmap went from 25% to 53% over the same period. Still 34 points behind adoption.
It's November. At many companies, the 2027 budget proposal is being closed now, not in January. What this data shows is that most legal departments will enter 2027 with more AI budget than they had in 2026, without having decided, with the same rigor, in what order that budget solves problems.
The challenge is no longer using AI. It's choosing where to use it first.
A survey published in March this year by Consilio, covering legal departments and law firms around the world, found something that reverses the logic of the last three years of talk about corporate AI. Choosing and implementing legal technology overtook workload as the biggest challenge reported by respondents. More than half cited tool selection and implementation as the central problem. Workload, the usual villain in this kind of survey, came in at 52%.
This doesn't mean the work has shrunk. It means the pain has moved. It's no longer about keeping up with demand. It's about not knowing which of ten possible workflows to automate will actually move the needle. The same FTI Consulting survey that showed 87% adoption also showed that around 70% of departments plan to budget for new legal technology in the next twelve months. Budget, then, is not the bottleneck. Criteria are.
Budget without criteria becomes adoption for adoption's sake
The practical risk of closing 2027 this way is familiar to any manager who has seen a new tool bought, rolled out and forgotten within six months. Without a roadmap that says which workflow goes first, the natural tendency is to invest in what is most visible, not in what weighs the most. Summarizing contracts is easier to demonstrate to a committee than measuring how much senior lawyer time is spent every month answering the same compliance question that has already been answered a hundred times before. The result is predictable. The company increases its AI investment and, a year later, still can't say with confidence what that investment solved.
Consider a mid-sized legal department that receives, by conservative market ranges, somewhere between 300 and 600 repetitive internal queries a month. An AI budget without defined priorities tends to automate what sits on the surface first, such as document triage or contract summaries, and leave aside what is hidden in the routine, such as the recurring question about internal policy that eats up one to two hours a week of each senior specialist's time. At the end of the year, the report shows AI adoption on the rise and the same volume of repetitive email landing in the same inbox. The budget grew. The invisible work did not shrink.
What changes when the budget comes with an order of priority
The difference between a department that increases its AI investment and solves something, and one that increases investment and merely keeps pace with the market average, is not the size of the check. It lies in deciding, before signing any new contract, which advisory domain concentrates the most Invisible Work and where automation should begin. That means mapping where the volume of repetitive questions is highest, how much of that volume currently goes through a specialist who could be doing something else, and which internal knowledge source would need to be organized before any AI could answer on its behalf safely.
That is exactly the principle behind Priority Sources, a feature AskLisa launched in 2026 that lets each area manager specify in advance which documents are the official reference for their domain before any automated answer is suggested. It isn't about switching AI on for everything at once. It's about deciding, area by area, what goes first, based on volume and risk, not visibility.
The 2027 budget is already being closed. The question that decides whether it's worth it isn't how much goes into it. It's whether someone, before signing, sat down to decide in what order it solves anything.

